N. C.
Doc.

J-£ NORTH CAROLINA STATE LIBRARY
40?/<3:S4 RALEIGH

" PHSB STUDIES —

A Special Report Series by the N,C. Department of Human Resources, Division of
Health Services, Public Health Statistics Branch, P.O. Box 2091, Raleigh, N.C.

No. ]k March 1979

ASSOCIATION OF INFLUENZA WITH CONGENITAL
TRACHEOESOPHAGEAL FISTULA AND OESOPHAGEAL ATRESIA:
AN ANALYSIS OF CLUSTERS

Epidemiologic Patterns of the Malformation

Research and management programs have had notable successes in solving many
problems associated with birth defects; however, the origin or causes of several defects
remain a mystery. Congenital tracheoesophageal fistula (TEF) and oesophageal atresia
(0A) are two such conditions whose etiology is not understood.

TEF is an abnormal passage between the trachea and the esophagus. It assumes any
of a large number of possible forms and is often characterized by a spectrum of symptoms
including choking, problems with swallowing and even death in a large percentage of
cases. 0A is a congenital lack of continuity of the esophagus and is characterized by
excessive salivation, gagging, vomiting, cyanosis and dyspnea. TEF and 0A often appear
together and are often studied together as though they were a single entity. We shall
follow this precedent if for no other reason than to amass enough cases.

Knox (1,2) presented some data that suggest that cases of TEF and 0A occur in
clusters over time. By considering the monthly and annual number of cases admitted to
hospitals in Birmingham, England from 1950 to 1955 and in the Newcastle region from
1950 to 1958, one finds years in which the incidence of reported cases was far in excess
of that of the surrounding years. However, it is less clear that a particular month
repeatedly was associated with a high reported rate. Babbott and Ingalls (3) found a
similar pattern in Pennsylvania County for the period 1951 through 1958, despite rela-
tively uniform birth rates. Koop (h) , in his experience with over 300 infants treated
for 0A in a 15_year period in Philadelphia, found that cases of 0A were admitted fairly
consistently over the years in bunches in April; this suggests a seasonal effect. He
reported that on one occasion in a period of 21 days, 17 infants with 0A were admitted
to the Neonatal Unit and that over half of these came from the same area of Pennsylvania
countryside. Other physicians have reported to us that the incidence of 0A and TEF seems
to be greater in April or May. Also, see the paper by Slater et al (5).

Contrary to this, there have been reports of the absence of clustering in time
or space of 0A. There are indeed times and places where clustering is absent but this
could simply reflect that a sporadically-behaved cause is absent as is true of infectious
diseases or it could imply an insensitive data-collecting mechanism. For example, we
have found that data from death certificates regarding certain malformations suggest
disparate conclusions about patterns of the disease depending on whether the condition
is listed as the underlying cause of death or simply mentioned at all.

Also, it is possible that a direct link may be found between a specific malfor-
mation as reflected in data comprised of underlying causes of death (with no or few
other conditions mentioned) and a specific environmental factor (e.g., influenza) and,
further, that this link may be obscured by including in the data deaths for which the
malformation is simply ment i oned . Conceivably, such a mention may be only a snail part
of a constellation of several anomalies associated with the infant, a constellation
caused by different factors (e.g., thalidomide, rubella).

There are perhaps other reasons why a disease that sometimes occurs in bunches
in a span of time at one place may at other places appear with some regularity. After
reviewing the little information that exists on the epidemiology of TEF and/or OA
(TEF/OA) and after comparing the patterns of TEF/OA with other congenital abnormalities,
we feel that clusters in time and space do occur and in this paper shall present some
statistical rationale to support this. The table below consists of incidence data of
births with TEF/OA as reported in investigations previously cited.

Predicted and Observed Cluster Sizes of Congenital
Tracheoesophageal Fistula and Oesophageal Atresia

Actual Maximum as Maximum
Consecutive Annual Cluster Predicted from Predicted if
Source Incidence Size Cluster Model No Clustering

BIRMINGHAM

1950-55 2, 2, 2, 15, 5, 9 15 15 <10

Knox

NEWCASTLE 1,8,1,4,3 8 9 6

1950-58 16, 11, 7, 12 16 2h 13

Knox

PENNSYLVANIA* 5,13,6,0 13 13 7

1951-58 7, 7, 2, 20 20 19 10

Babbott, Ingalls

*These data represent only TEF.

Grimson (6) has developed a cluster model that appears appropriate, both in
terms of rationale and fit, in describing infectious processes. By considering the
span of years to be h, 5 or 6 (this is a parameter in the model called the cluster
modulus, which is similar to the idea of periodicity), the actual maximum cluster size
is, with good accuracy, predicted by the cluster model. In general, it significantly
exceeds the estimates from the Ederer-Myers-Mantel model which produces maximum numbers
under the assumption of no clustering; this provides a statistical test for clustering
(7,8).

Viral etiology has been implicated in some anomalies; in particular, it is widely
accepted that rubella, maternal cytomegalovirus and coxsackieviruses act on the fetus.
Other infectious diseases including hepatitis, mumps and influenza have been suggested
as correlates of congenital anomalies but little or no hard evidence exists, especially
in connection with TEF/OA.

Few epidemiological attempts have been made to try to link influenza with TEF,
0A and other anomalies. In addition to providing a quantitative logic and description
behind the cluster question, we shall offer a conjecture that women who are in their
first trimester of pregnancy during a period of a high incidence of influenza A are at
higher risk of giving birth to a child with TEF/OA than are women in early stages
of pregnancy at other times.

Associations

In noting the non-random patterns of TEF/OA, we decided to map on a time axis
the incidence in North Carolina of last menstrual periods plus a month for mothers
giving birth to a TEF/OA child who lived for less than one year and whose underlying
cause of death was TEF/OA as determined from the ICDA codes (8th revision) 750.2 and
750.3. Mortality statistics are used because they can be obtained from existing records.
TEF/OA was not associated with fetal death according to fetal death certificates. Live
birth certificates did not, until recently, provide a place for listing malformations.
It is reasonable that we restrict ourselves to a one-year life span because, beyond that,
TEF/OA is rarely given as an underlying cause of death; also, this permits us to include
all deaths corresponding to births conceived during the ten-year period 1967-1976.
Unfortunately, comparable data for earlier deaths are not available.

We excluded those records in which TEF/OA was mentioned but the underlying cause
of death was something else. These cases tended to have multiple anomalies whose
etiology may be different from those that are specific to the respiratory and digestive
tract area. Our hypothes i s is that influenza is associated specifically or most imme-
diately with TEF/OA and not necessarily with multiple anomalies.

The results are represented in the display on page k. Drug prescription trends,
changes in the birth rate over time, changes in age patterns of pregnant women, improved
life-saving treatment for TEF/OA (not thought to be significant since 1967), patterns of
rubella and other factors may correlate with the patterns observed here but probably
none match as closely as influenza patterns. Rubella and TEF/OA are not correlated in
these data. Mortality due to influenza reflects morbidity, so annual influenza mortality
figures were used.

The correlation between the last menstrual period for TEF/OA infant deaths and
influenza deaths is striking; the Hotel 1 ing-Pabst test (9) shows that the correlation is
different from zero at the .025 a level. The mortality due to TEF/OA is a subset of
the total number of cases so conclusions assume that mortality patterns reflect the
incidence. In support of the pattern, we do note that the case fatality rate is high.

Another bit of evidence of the association is the fact that the large number of
last menstrual periods that occurred during the 1 968-69 winter corresponds exactly with
the dramatic A/Hong Kong influenza epidemic. The reversals in the order of magnitudes
of TEF/OA data and the influenza data in 1 968 and 1969 could be explained by deaths
occurring more often at the later stages of the epidemic. The year 1970 was another
memorable influenza year in North Carolina and, again, there appears a cluster of TEF/OA
deaths.

A corresponding bit of evidence can be seen in the data of Babbott and Ingalls (3);
see our first table. Although these authors did not mention it, the large cluster of
TEF cases occurring in 1958 would be in early fetal developmental stages during the
major influenza A2 (Asian) pandemic of October 1957 through March 1958. In an earlier
study that was also based on small numbers and seems to have been overlooked, Leek (10)
suggests that the incidence of TEF/OA between 26 and 40 weeks following influenza
epidemics is over twice as large than at other times.

Concl us ion

We have provided a mathematical description of the ability of TEF/OA to cluster
in time. Also, we have introduced new evidence and reinterpreted existing data that
suggest that women who are within early stages of pregnancy during times of high
incidence of Type A influenza are at higher risk of giving birth to a child with TEF/OA
than are women who are in early stages of pregnancy at other times.

As illustrated here, vital statistics data form an important part, if not the
key part, in the formulation and/or support of medical hypotheses. Here is a unique
large source of data and an elaborate data collecting process that can be used profit-
ably by investigators or planners who understand the features of the data.

As a follow-up to the present study, we have begun to review charts at North
Carolina Memorial Hospital in hopes of establishing a similar pattern for live patients
and, eventually, for all patients in North Carolina over several years. Also, thoughts
are being given to the prospect of a case-control study.

TRENDS IN TEF/OA AND INFLUENZA MORTALITY

1967 1968 1969 1970 1971 1972 1973 1974 1975 1976

|..,.T...T.T.|.TT..T?..TTl|rT\.T ff.Ttft I...T I. .T.T...T...I....T....T..I If T,.,| I

<
O

u_

UJ

3

8

4

7

<

N

0

42

□

178

283

209

209

23

162

121

64

56

Monthly patterns of last menstrual periods (plus a month) of motners
who gave birth to a child with TEF/OA, as determined from North
Carolina resident infant death certificates. A correlation with
annual influenza mortality in North Carolina is observed.

REFERENCES

(1) Knox, G. "Secular Pattern of Congenital Oesophageal Atresia." Brit. J. Prev.

Soc. Med. , Vol. 13, pp. 222-226, 1959-

(2) Knox, E. G. "Epidemics of Rare Diseases." Brit. Med. Bull. , Vol. 27, No. 1,

PP. ^3-^7, 1971.

(3) Babbott, J. G. and Ingalls, T. H. "Tracheoesophageal Fistula Occurring in

Pennsylvania." Quarterly Review of Pediatrics, Vol. 16, pp. 86-92, April-
June, 1961.

(4) Koop, C. E. "Recent Advances in the Surgery of Oesophageal Atresia." Progress

in Pediatric Surgery, Vol. 2, pp. 41-56, 1971.

(5) Slater, B. C. S. , Watson, G. I. and McDonald, J. C. "Seasonal Variation in

Congenital Abnormalities." Brit. J. Prev. Soc. Med. , Vol. 18, pp. 107, 196*4.

(6) Grimson, R. C. "A Cluster Model and a New Characteristic of the Epidemicity of

Hepatitis A." To appear.

(7) Ederer, F. , Myers, Max H. and Mantel, N. "A Statistical Problem in Space and

Time: Do Leukemia Cases Come in Clusters?" Biometrics , Vol. 20, No. 3,
pp. 626-638, September 1 964.

(8) Mantel, N., Kryscio, R. J. and Myers, M. H. "Tables and Formulas for Extended

Use of the Ederer-Myers-Mantel Disease-Clustering Procedure." American
Journal of Epidemiology, Vol. 104, No. 5, pp. 576-584, 1976.

(9) Conover, W. J. Practical Nonparametric Statistics, John Wiley & Sons Inc., New

York, 1971.

(10) Leek, I. "Incidence of Malformations Following Influenza Epidemics." Brit. J.
Prev. Soc. Med. , Vol. 17, pp. 70-80, 1963.

A major portion of the work for this paper and the
paper itself are attributable to Dr. Roger Grimson,
statistical consultant from the Department of Bio-
statistics, School of Public Health, University of
North Carol ina.

STATE LIBRARY OF NORTH CAROLINA

3 3091 00747 2483

Public Health Statistics Branch
Division of Health Services
Department of Human Resources
P.O. Box 209 l
Raleigh, North Carolina 27602